<?xml version="1.0" standalone="yes"?>
<Paper uid="C82-2044">
  <Title>Miusk~, Marvin 1975. &amp;quot;A Framework for Representing Knowledge&amp;quot; in Wins~n (ed.), The Ps~oholo~ of Computer Vision.</Title>
  <Section position="1" start_page="0" end_page="0" type="abstr">
    <SectionTitle>
MEANING NEGOTIATION IN DIALOGUE
</SectionTitle>
    <Paragraph position="0"> For the sake of effective conmmication of man-machine type, what seems necessary is to find means to deal with complex language problems in ways that would guarantee its correct analysis and synthesis. Such problems can be better understood when language is viewed not as a static product of interaction but as a dynamic, meaning producing process in action. The importance of looking at ongoing conmmication in order to achieve a fuller understanding of the numerous linguistic and nonlinguistic devices employed in the process of human communication has been recently emphasized by manyauthors.</Paragraph>
    <Paragraph position="1"> (Clark and Clark'68, Hallday an~ Hasan &amp;quot;76, Garfinkel &amp;quot;72, Churchill &amp;quot;78) * The meaning of an utterance: can be considered to be a complex that consists of the semantic level at the assumption of some idealization of data, the pragmatic level, elicited by the context and represented as a set of rules of language use, as well as the material referring to the knowledge of the world.</Paragraph>
    <Paragraph position="2"> A correct analysis of language texts should make it possible to perform the closest recovery both of what was said and of what was talked about. The information required to interpret and reproduce an utterance is not fully and unambiguously encoded in the speech signal, but is completed by the material contained in sets of constraints familiar to each of the interlocutors. If the relations between the signal and the</Paragraph>
    <Paragraph position="4"> content correspond to generally accepted rules (syntactic, semantic, pra~atic), furthermore, if the relations intended by the sender are identical to those apprehended by the receiver, the co~nunication is conventional (Allwooddeg76). Conversation, and most explicitly, dialogue, is a continuous recovering of the above relations. The issue of whether those factors are entirely or partly identical or different to the participants is being established in the process of the whole interaction. Hence, meaning of an utterance is not a constant but a variable the value of which is negotiated in the course of interaction. Applying Ninsky &amp;quot;s terms (~insk~7 &amp;quot;75) what takes place there is the process of filling in &amp;quot;default&amp;quot;values in prototypes, u~derstood here as hiersrchically organized data structures, representing partial but &amp;quot;constant&amp;quot; knowledge, with the variable defaults, empty, prior to the concrete act of perception. This, in turn, leads either to activating familiar conventional or stereotypical conceptual frames or stimulates instantiating new data structures if the higher &amp;quot;constant&amp;quot; nodes in the structure are replaced. A number of man-machine communication systems (FRL-Roberts, Goldstein, KRL - Bobrow, Winograd, Norman, Kay - cf. Jirk~ &amp;quot;81 for details) work according to similar schemes, which makes it possible for them to include sets of inference rules in their systems.</Paragraph>
    <Paragraph position="5"> Inference rules underlying human as well as man-machine co~nmication can be represented in terms of data structure reconfigurations according to known typological criteria such as hyponymy, quantification, deletions, additions, etc. Such changes can be perceived as opaque by participants due to the lack of transparency in the linguistic form of the utterance.</Paragraph>
    <Paragraph position="6"> Another problem in this connection is that even in the simplest forms of dialogue, question answering, the chain maxim, i.e. respondlng with a direct answer to the question, is - 183 strictly followed in only a small percentage of cases (12 24 % of oases - Churchill op. oit.). Other responses exhibit a number of forms andpatterns rangin8 from most conventional (conforming) to most devtant. Deviations from the direct patterns can be hierarcb/cally classified according to breaches of conversational maxims.</Paragraph>
    <Paragraph position="7"> In order then, to correctly analyse such and similar dialogues as the following (after Churchil op.oit.; 103): Speaker1: rny are we goingw~7 out in the middle? I'll get sunburned Speaker2: What's the difference whether you're in the middle or not? Speaker1: You get more reflection in the middle Speaker2: Oh.</Paragraph>
    <Paragraph position="8"> A system is proposed in the present paper, oomb~aiz~ differ=.</Paragraph>
    <Paragraph position="9"> ent typological orlteriaunderlying inferences in frame terms with hierarchical patterning of conversational deviations.</Paragraph>
  </Section>
class="xml-element"></Paper>